Subject: qube2700, 3C905b/c : first step & success
To: None <port-cobalt@NetBSD.org>
From: Axel Morhenn <morhenn@comsult.net>
List: port-cobalt
Date: 02/02/2004 00:13:16
Hi @ll,

first step done. I was able to do a netboot with the qube2700 and get =
access
via telnet :) it works and the Qube2700 is able to run NetBSD at all.

at the moment, i use a simple modifyed 161-release-kernel cross-compiled
from source. The bigest problem was the init-process in the netbsd 161; =
the
init-daemon checks for /dev/console.=20

If it finds /dev/console, an error-message will be generated and the
boot-process stops. if there is no /dev/console the system try to use
/dev/MAKEDEV to generate the missing dev-links. but (why at hell ? ) the
process fails and the system is hanging again.=20

I made the inetd (with telnet enabled) starting from /dev/MAKEDEV (and =
then
an 'exit 0'). Ok, ok, quick and very very very dirty. The kernel must =
have
'com0 at mainbus?' defined (otherwise telnet will terminated during the
login); i did comment out=20
//      com_attach_subr(sc);                                             =
  =20
//      cpu_intr_establish(maa->ma_level, IPL_SERIAL, comintr, sc);      =
  =20

in arch/cobalt/dev/com_mainbus.c ; otherwise the kernel-boot failed.

after telnet into the box i was able to run the install-skript from the
netbsd-rescue-images; everything went fine.=20

until now i was not able to boot from disc; i think there is a problem =
with
the partition-table for the rom-booting. it seems that the rom-boot does =
not
find the kernel or boot-partition. ok, it will work with a litle more =
time i
hope <g>

@cdi: i was not able to use your console.c-hack; could you please give =
me a
short step-by-step-description how to patch a 'normal' =
release-kernel-source
(without any other patches e.g. lcd.diff or hdxxxx.diff) to that level =
using
the lcd as console ? maybe the init-problem is solved by that.

@all
i am 'brand-new' at netbsd, therefor i have too less knowledge of the =
161
init-processes. i hope, somebody could help  finding a way to eliminate =
the
init/console-problem for the 2700er.

 ok, thats the situation at the moment.=20

so long,

axel
                                                                         =
  =20